Retinex增强---MATLAB

Retinex增强---MATLAB

1、retinex_frankle_mccann.m代码

function Retinex = retinex_frankle_mccann(L, nIterations)

% RETINEX_FRANKLE_McCANN: 
%         Computes the raw Retinex output from an intensity image, based on the
%         original model described in:
%         Frankle, J. and McCann, J., "Method and Apparatus for Lightness Imaging"
%         US Patent #4,384,336, May 17, 1983
%
% INPUT:  L           - logarithmic single-channel intensity image to be processed
%         nIterations - number of Retinex iterations
%
% OUTPUT: Retinex     - raw Retinex output
%
% NOTES:  - The input image is assumed to be logarithmic and in the range [0..1]
%         - To obtain the retinex "sensation" prediction, a look-up-table needs to
%         be applied to the raw retinex output
%         - For colour images, apply the algorithm individually for each channel
%
% AUTHORS: Florian Ciurea, Brian Funt and John McCann. 
%          Code developed at Simon Fraser University.
%
% For information about the code see: Brian Funt, Florian Ciurea, and John McCann
% "Retinex in Matlab," by Proceedings of the IS&T/SID Eighth Color Imaging 
% Conference: Color Science, Systems and Applications, 2000, pp 112-121.
%
% paper available online at http://www.cs.sfu.ca/~colour/publications/IST-2000/
%
% Copyright 2000. Permission granted to use and copy the code for research and 
% educational purposes only.  Sale of the code is not permitted. The code may be 
% redistributed so long as the original source and authors are cited.

global RR IP OP NP Maximum
RR = L;
Maximum = max(L(:));                                 % maximum color value in the image
[nrows, ncols] = size(L);

shift = 2^(fix(log2(min(nrows, ncols)))-1);          % initial shift
OP = Maximum*ones(nrows, ncols);                     % initialize Old Product

while (abs(shift) >= 1)
   for i = 1:nIterations
      CompareWith(0, shift);                         % horizontal step
      CompareWith(shift, 0);                         % vertical step
   end
   shift = -shift/2;                                 % update the shift
end
Retinex = NP;

function CompareWith(s_row, s_col)
global RR IP OP NP Maximum
IP = OP;
if (s_row + s_col > 0)
   IP((s_row+1):end, (s_col+1):end) = OP(1:(end-s_row), 1:(end-s_col)) + ...
   RR((s_row+1):end, (s_col+1):end) - RR(1:(end-s_row), 1:(end-s_col));
else
   IP(1:(end+s_row), 1:(end+s_col)) = OP((1-s_row):end, (1-s_col):end) + ...
   RR(1:(end+s_row),1:(end+s_col)) - RR((1-s_row):end, (1-s_col):end);
end
IP(IP > Maximum) = Maximum;                          % The Reset operation
NP = (IP + OP)/2;                                    % average with the previous Old Product
OP = NP;                                             % get ready for the next comparison

2、 retinex_mccann99.m代码

function Retinex = retinex_mccann99(L, nIterations)

% RETINEX_McCANN99: 
%         Computes the raw Retinex output from an intensity image, based on the
%         more recent model described in:
%         McCann, J., "Lessons Learned from Mondrians Applied to Real Images and 
%         Color Gamuts", Proc. IS&T/SID Seventh Color Imaging Conference, pp. 1-8, 1999
%
% INPUT:  L           - logarithmic single-channel intensity image to be processed
%         nIterations - number of Retinex iterations
%
% OUTPUT: Retinex     - raw Retinex output
%
% NOTES:  - The input image is assumed to be logarithmic and in the range [0..1]
%         - To obtain the retinex "sensation" prediction, a look-up-table needs to
%         be applied to the raw retinex output
%         - For colour images, apply the algorithm individually for each channel
%
% AUTHORS: Florian Ciurea, Brian Funt and John McCann. 
%          Code developed at Simon Fraser University.
%
% For information about the code see: Brian Funt, Florian Ciurea, and John McCann
% "Retinex in Matlab," by Proceedings of the IS&T/SID Eighth Color Imaging 
% Conference: Color Science, Systems and Applications, 2000, pp 112-121.
%
% paper available online at http://www.cs.sfu.ca/~colour/publications/IST-2000/
%
% Copyright 2000. Permission granted to use and copy the code for research and 
% educational purposes only.  Sale of the code is not permitted. The code may be 
% redistributed so long as the original source and authors are cited.

global OPE RRE Maximum
[nrows ncols] = size(L);                             % get size of the input image
nLayers = ComputeLayers(nrows, ncols);               % compute the number of pyramid layers
nrows = nrows/(2^nLayers);                           % size of image to process for layer 0
ncols = ncols/(2^nLayers);
if (nrows*ncols > 25)                                % not processing images of area > 25
  error('invalid image size.')                       % at first layer
end
Maximum = max(L(:));                                 % maximum color value in the image
OP = Maximum*ones([nrows ncols]);                    % initialize Old Product
for layer = 0:nLayers
   RR = ImageDownResolution(L, 2^(nLayers-layer));   % reduce input to required layer size
   
   OPE = [zeros(nrows,1) OP zeros(nrows,1)];         % pad OP with additional columns
   OPE = [zeros(1,ncols+2); OPE; zeros(1,ncols+2)];  % and rows
   RRE = [RR(:,1) RR RR(:,end)];                     % pad RR with additional columns
   RRE = [RRE(1,:); RRE; RRE(end,:)];                % and rows
   
   for iter = 1:nIterations
     CompareWithNeighbor(-1, 0);                     % North
     CompareWithNeighbor(-1, 1);                     % North-East
     CompareWithNeighbor(0, 1);                      % East
     CompareWithNeighbor(1, 1);                      % South-East
     CompareWithNeighbor(1, 0);                      % South
     CompareWithNeighbor(1, -1);                     % South-West
     CompareWithNeighbor(0, -1);                     % West
     CompareWithNeighbor(-1, -1);                    % North-West
   end
   NP = OPE(2:(end-1), 2:(end-1));
   OP = NP(:, [fix(1:0.5:ncols) ncols]);             %%% these two lines are equivalent with 
   OP = OP([fix(1:0.5:nrows) nrows], :);             %%% OP = imresize(NP, 2) if using Image
   nrows = 2*nrows; ncols = 2*ncols;                 % Processing Toolbox in MATLAB
end
Retinex = NP;

function CompareWithNeighbor(dif_row, dif_col)
global OPE RRE Maximum

% Ratio-Product operation
IP = OPE(2+dif_row:(end-1+dif_row), 2+dif_col:(end-1+dif_col)) + ...
     RRE(2:(end-1),2:(end-1)) - RRE(2+dif_row:(end-1+dif_row), 2+dif_col:(end-1+dif_col));
     
IP(IP > Maximum) = Maximum;                          % The Reset step

% ignore the results obtained in the rows or columns for which the neighbors are undefined
if (dif_col == -1) IP(:,1) = OPE(2:(end-1),2); end
if (dif_col == +1) IP(:,end) = OPE(2:(end-1),end-1); end
if (dif_row == -1) IP(1,:) = OPE(2, 2:(end-1)); end
if (dif_row == +1) IP(end,:) = OPE(end-1, 2:(end-1)); end
NP = (OPE(2:(end-1),2:(end-1)) + IP)/2;              % The Averaging operation
OPE(2:(end-1), 2:(end-1)) = NP;

function Layers = ComputeLayers(nrows, ncols)
power = 2^fix(log2(gcd(nrows, ncols)));              % start from the Greatest Common Divisor
while(power > 1 & ((rem(nrows, power) ~= 0) | (rem(ncols, power) ~= 0)))
   power = power/2;                                  % and find the greatest common divisor
end                                                  % that is a power of 2
Layers = log2(power);

function Result = ImageDownResolution(A, blocksize)
[rows, cols] = size(A);                              % the input matrix A is viewed as
result_rows = rows/blocksize;                        % a series of square blocks
result_cols = cols/blocksize;                        % of size = blocksize
Result = zeros([result_rows result_cols]);
for crt_row = 1:result_rows                          % then each pixel is computed as
   for crt_col = 1:result_cols                       % the average of each such block
      Result(crt_row, crt_col) = mean2(A(1+(crt_row-1)*blocksize:crt_row*blocksize, ...
                                       1+(crt_col-1)*blocksize:crt_col*blocksize));
   end
end

3、实例-调用1

Test.m代码：

src=imread('fig2.tif');  
[m,n,z] = size(src);  
dst = zeros(m,n,z);  
for i = 1:z  
    ImChannel = log(double(src(:,:,i))+eps);  
    dst(:,:,i)=retinex_mccann99(ImChannel,4);   
    dst(:,:,i)=exp(dst(:,:,i));  
    a=min(min(dst(:,:,i)));  
    b=max(max(dst(:,:,i)));  
    dst(:,:,i)=((dst(:,:,i)-a)/(b-a))*255;       
end  
dst=uint8(dst);  
figure(1);  
imshow(src);  
figure(2);  
imshow(dst);  
imwrite(dst,'result.tif');

运行结果

实例2

Test.m为：

clc;
clear all;
close all;
src=imread('fig2.tif');  
[m,n,z] = size(src);  
dst = zeros(m,n,z);  
for i = 1:z  
    ImChannel = log(double(src(:,:,i))+eps);  
    dst(:,:,i)=retinex_frankle_mccann(ImChannel,4);   
    dst(:,:,i)=exp(dst(:,:,i));  
    a=min(min(dst(:,:,i)));  
    b=max(max(dst(:,:,i)));  
    dst(:,:,i)=((dst(:,:,i)-a)/(b-a))*255;       
end  
dst=uint8(dst);  
figure(1);  
imshow(src);  
figure(2);  
imshow(dst);  
imwrite(dst,'result.tif');

运行结果

另一图片运行结果为：